People tracking and identification with a mobile robot

In this paper we present a novel and efficient solution for tracking and identifying people with a mobile robot using multisensor data fusion. The system utilizes a laser device to detect human legs and a PTZ camera to find faces, then the relative data is fused with a sequential Unscented Kalman Filter to perform real-time tracking. A metric based on the Bhattacharyya coefficient for color histogram comparison is also adopted to identify persons wearing different clothes. Finally, integrating the information coming from the tracking and the identification modules, we improve the robustness of the data association process. Some experiments with a mobile robot show the effectiveness of our approach.